Adaptive robust bidding method for virtual power plant

Abstract

The embodiment of the invention relates to the field of data processing, in particular to an adaptive robust bidding method and system for a virtual power plant. The method comprises the steps that a load fuzzy response mechanism model is adopted, a load transfer parameter is generated according to time-of-use electricity price obtained in real time, and a bidding basis is obtained based on the load transfer parameter; obtaining historical transaction data of the first transaction day, and predicting first prediction bidding strategy data in the second transaction day based on the historical transaction data through the day-ahead market bidding strategy model; performing real-time bidding strategy calculation through a real-time market bidding strategy model based on the real-time transaction data of the day of the second transaction day to obtain second prediction bidding strategy data; and dynamically adjusting the real-time electricity price transaction information in each time period based on the second prediction bidding strategy data so as to realize a self-adaptive robust electric power bidding strategy. And through a load fuzzy response mechanism and an adaptive robust optimization model, bidding risks are reduced, and the economic benefits of the virtual power plant and the stability of the power market are improved.

Description

Technical Field

[0001] Embodiments of the present application relate to the field of data processing, and in particular to an adaptive robust bidding method and system for a virtual power plant. Background Art

[0002] Among the related technologies, the power sector has made significant progress in the application of renewable energy due to its environmentally friendly characteristics and global energy shortage. However, the randomness and volatility of wind and photovoltaic power generation pose challenges to the safe and stable operation of the power system, especially in virtual power plants, where the integration of different types of distributed energy resources faces technical barriers. The two main modeling methods for dealing with the uncertainty of renewable energy output are stochastic optimization (SO) and robust optimization (RO). SO captures uncertainty through statistical models, but the calculations are complex and the samples may be inaccurate. RO searches for the optimal solution without predetermining the probability distribution, but is often criticized for being too conservative.

[0003] In response to the above problems, in related technologies, data-driven uncertainty sets based on historical data are used. For example, some scholars use Bayesian non-parametric methods to mine the spatiotemporal characteristics of historical weather and grid information to develop uncertainty sets. However, this method relies on the selection of prior probabilities, which may introduce inaccuracies and affect the posterior probability distribution. At the same time, this method is computationally complex. Although artificial intelligence and kernel density estimation (KDE) methods provide multi-step probability density predictions, they also face problems in kernel function and bandwidth selection, and are easily affected by data distribution, resulting in errors in the estimation results.

[0004] In summary, it is urgent to propose a completely new technical solution to solve at least one of the above technical problems. Summary of the invention

[0005] In response to the technical problems existing in the prior art, the embodiments of the present application provide an adaptive robust bidding method and system for a virtual power plant, which is used to dynamically adjust the bidding strategy based on time-of-use electricity prices through a load fuzzy response mechanism and an adaptive robust optimization model, so as to effectively cope with electricity price fluctuations and uncertainty in renewable energy output; optimize bidding plans using historical transaction data and real-time data, reduce bidding risks, and improve the economic benefits of virtual power plants and the stability of the electricity market.

[0006] In a first aspect, an embodiment of the present application provides an adaptive robust bidding method and system for a virtual power plant, the method comprising:

[0007] A load fuzzy response mechanism model is adopted to generate load transfer parameters according to the time-of-use electricity price obtained in real time, and a bidding basis is obtained based on the load transfer parameters; the bidding basis is used to indicate the total amount of bidding in multiple markets;

[0008] Obtain historical transaction data of the first trading day, and predict first predicted bidding strategy data for the second trading day based on the historical transaction data through a day-ahead market bidding strategy model; the historical transaction data includes at least one of the following: historical wind power output data, historical photovoltaic output data, historical market electricity price data, and historical user load change data; the first predicted bidding strategy data includes at least one of the following: gas turbine output, energy storage device power, and output of transferable and interruptible load projects;

[0009] Based on the real-time transaction data of the second trading day, a real-time bidding strategy is calculated by a real-time market bidding strategy model to obtain second predicted bidding strategy data for adjusting the real-time bidding power in the bidding basis; the second predicted transaction data includes at least one of the following: electric vehicle charging and discharging power, multi-market bidding output;

[0010] Based on the second predicted bidding strategy data, the real-time electricity price transaction information in each time period in the first predicted bidding strategy data is dynamically adjusted to achieve an adaptive and robust electricity bidding strategy.

[0011] In a second aspect, an embodiment of the present application provides an adaptive robust bidding system for a virtual power plant, the system comprising the following units, wherein:

[0012] A generating unit is configured to adopt a load fuzzy response mechanism model, generate a load transfer parameter according to a time-of-use electricity price obtained in real time, and obtain a bidding basis based on the load transfer parameter; the bidding basis is used to indicate the total amount of bidding to participate in multiple markets;

[0013] The first prediction unit is configured to obtain historical transaction data of the first trading day, and predict first predicted bidding strategy data within the second trading day based on the historical transaction data through a day-ahead market bidding strategy model; the historical transaction data includes at least one of the following: historical wind power output data, historical photovoltaic output data, historical market electricity price data, and historical user load change data; the first predicted bidding strategy data includes at least one of the following: gas turbine output, energy storage device power, and output of transferable and interruptible load projects;

[0014] The second prediction unit is configured to perform a real-time bidding strategy calculation based on the real-time transaction data of the second trading day through a real-time market bidding strategy model to obtain second predicted bidding strategy data for adjusting the real-time bidding power; the second predicted transaction data includes at least one of the following: electric vehicle charging and discharging power, and multi-market bidding output;

[0015] The real-time bidding unit is configured to dynamically adjust the real-time electricity price transaction information in each time period in the first predictive bidding strategy data based on the second predictive bidding strategy data to achieve an adaptive and robust electricity bidding strategy.

[0016] In a third aspect, an embodiment of the present application provides an electronic device, the electronic device comprising:

[0017] at least one processor, memory, and input-output unit;

[0018] The memory is used to store a computer program, and the processor is used to call the computer program stored in the memory to execute the adaptive robust bidding method for the virtual power plant of the first aspect.

[0019] In a fourth aspect, a computer-readable storage medium is provided, comprising instructions, which, when executed on a computer, enable the computer to execute the adaptive robust bidding method for a virtual power plant of the first aspect.

[0020] The beneficial effects of the embodiments of the present application are: providing an adaptive robust bidding method and system for a virtual power plant. In this technical solution, a load fuzzy response mechanism model is adopted to generate load transfer parameters according to the time-of-use electricity price obtained in real time, and a bidding basis is obtained based on the load transfer parameters; the historical transaction data of the first trading day is obtained, and the first predicted bidding strategy data in the second trading day is predicted based on the historical transaction data through the day-ahead market bidding strategy model; based on the real-time transaction data of the second trading day, the real-time bidding strategy is calculated through the real-time market bidding strategy model to obtain the second predicted bidding strategy data for adjusting the real-time bidding electricity in the bidding basis; based on the second predicted bidding strategy data, the real-time electricity price transaction information in each time period in the first predicted bidding strategy data is dynamically adjusted to realize an adaptive robust electricity bidding strategy. This technical solution uses a load fuzzy response mechanism and an adaptive robust optimization model to dynamically adjust the bidding strategy based on the time-of-use electricity price, effectively responding to electricity price fluctuations and uncertainty in renewable energy output; historical transaction data and real-time data are used to optimize the bidding scheme, reduce bidding risks, and improve the economic benefits of virtual power plants and the stability of the electricity market. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flowchart of an adaptive robust bidding method for a virtual power plant according to an embodiment of the present application;

[0022] Figures 2 to 8 It is a schematic diagram of the principle of an adaptive robust bidding method for a virtual power plant in an embodiment of the present application;

[0023] Fig. 9It is a structural schematic diagram of an adaptive robust bidding system for a virtual power plant according to an embodiment of the present application;

[0024] Fig.10 is a schematic diagram of the structure of an electronic device according to an embodiment of the present application;

[0025] Fig.11 It is a structural schematic diagram of a medium device according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0027] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, "plurality" means two or more, unless otherwise clearly and specifically defined.

[0028] In the description of the present application, the term "for example" is used to mean "used as an example, illustration or description". Any embodiment described as "for example" in the present application is not necessarily to be interpreted as being more preferred or more advantageous than other embodiments. In order to enable any technician in the field to implement and use the embodiments of the present application, the following description is given. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the embodiments of the present application can also be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid unnecessary details that make the description of the embodiments of the present application obscure. Therefore, the embodiments of the present application are not intended to be limited to the embodiments shown, but are consistent with the widest range of principles and features disclosed in the present application.

[0029] The power sector has made great progress in implementing renewable energy sources worldwide due to their environmentally friendly nature and the energy shortage crisis. The penetration of renewable energy in the distribution system has reached a considerable level due to the progress of economic, technological, social and environmental policies. Although renewable energy sources such as wind and photovoltaic power generation can bring huge benefits to the power system, the randomness and volatility of their output due to the lack of participation in the liberalized electricity market is one of the most important challenges facing the safe and stable operation of the power system. For example, in virtual power plants (VPPs), the integration of different types of distributed energy resources (DER), energy storage and adjustable loads still faces technical challenges, especially when coordinating them to provide stability and resilience while ensuring economic feasibility.

[0030] In the related art, the two main modeling approaches to deal with the challenges of managing the volatility and uncertainty of renewable energy output are stochastic optimization (SO) and robust optimization (RO). SO technology models uncertainty through statistical models. In order to accurately reflect the probability distribution of uncertainty, scenario-based modeling, Monte Carlo simulation, and point estimation methods have been widely used. Although the model fully represents time-dependent quantities, it brings a considerable computational burden, and it cannot guarantee that the construction of scenario samples is accurate enough to represent future trends. Compared with SO, RO can obtain the optimal solution to uncertainty without the need to obtain the precise probability distribution in advance, but this solution is often criticized as being too conservative and ignoring the potential economic value of the two-stage RO plan. At present, the adaptive robust optimization framework has been widely used in power systems to cope with the impact of uncertainty brought by renewable energy output and load demand, such as unit commitment, economic dispatch, and reactive power optimization. These studies discuss various methods for constructing uncertainty sets, including cuboid uncertainty sets, ellipsoid uncertainty sets, polyhedron uncertainty sets, and budget uncertainty sets. However, it requires additional efforts to determine the appropriate boundaries.

[0031] In order to solve the overly conservative nature of the above uncertainty sets, data-driven uncertainty sets based on historical data have been proposed in recent years. For example, some scholars have used Bayesian non-parametric methods to mine the spatiotemporal characteristics of historical weather and grid information to develop uncertainty sets. However, this method relies on the selection of prior probabilities, which may introduce inaccuracies and affect the posterior probability distribution. In addition, calculating the posterior probability is time-consuming and difficult. Alternatively, some scholars have used artificial intelligence algorithms as a basis, supplemented by kernel density estimation (KDE) methods to achieve multi-step probability density predictions for photovoltaic and wind power generation. However, KDE requires specifying a fixed kernel function and bandwidth for density estimation, which is easily affected by the unevenness and diversity of data distribution and produces errors in the estimation results.

[0032] It is urgent to propose a completely new technical solution to solve at least one of the above technical problems.

[0033] The embodiment of the present application provides an adaptive robust bidding method and system for a virtual power plant. In this technical solution, a load fuzzy response mechanism model is adopted to generate load transfer parameters according to the time-of-use electricity price obtained in real time, and a bidding basis is obtained based on the load transfer parameters; the bidding basis is used to indicate the total amount of bidding in multiple markets; the historical transaction data of the first trading day is obtained, and the first predicted bidding strategy data within the second trading day is predicted based on the historical transaction data through the day-ahead market bidding strategy model; the historical transaction data includes at least one of the following: historical wind power output data, historical photovoltaic output data, historical market electricity price data, historical user load change data; the first predicted bidding strategy data includes at least The real-time bidding strategy model is used to calculate the real-time bidding strategy based on the real-time trading data of the second trading day, and the second predicted bidding strategy data is used to adjust the real-time bidding power in the bidding basis; the second predicted trading data includes at least one of the following: electric vehicle charging and discharging power, multi-market bidding output; based on the second predicted bidding strategy data, the real-time electricity price trading information in each time period in the first predicted bidding strategy data is dynamically adjusted to achieve an adaptive and robust electricity bidding strategy.

[0034] From a technical perspective, in the embodiments of the present application, through the load fuzzy response mechanism and the adaptive robust optimization model, the bidding strategy is dynamically adjusted based on the time-of-use electricity price to effectively respond to electricity price fluctuations and the uncertainty of renewable energy output; historical transaction data and real-time data are used to optimize the bidding plan, reduce bidding risks, and improve the economic benefits of virtual power plants and the stability of the electricity market.

[0035] The adaptive robust bidding scheme for a virtual power plant provided in an embodiment of the present application may also be executed by an electronic device, which may be a server, a server cluster, or a cloud server. The electronic device may also be a terminal device such as a mobile phone, a computer, a tablet computer, a wearable device, or a dedicated device (such as a dedicated terminal device with an adaptive robust bidding system for a virtual power plant, etc.). These electronic devices may also be equipped with the chips described in the above embodiments. Alternatively, these electronic devices may also be installed with a service program for executing an adaptive robust bidding scheme for a virtual power plant.

[0036] Figure 1 A flowchart of an adaptive robust bidding method for a virtual power plant provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the method comprises the following steps:

[0037] 101, using a load fuzzy response mechanism model, generating a load transfer parameter according to the time-of-use electricity price obtained in real time, and obtaining a bidding basis based on the load transfer parameter;

[0038] 102, obtaining historical transaction data of the first trading day, and predicting first predicted bidding strategy data for the second trading day based on the historical transaction data through a day-ahead market bidding strategy model;

[0039] 103, based on the real-time transaction data on the second trading day, a real-time bidding strategy calculation is performed through a real-time market bidding strategy model to obtain second predicted bidding strategy data for adjusting the real-time bidding power in the bidding basis;

[0040] 104. Based on the second predicted bidding strategy data, dynamically adjust the real-time electricity price transaction information in each time period in the first predicted bidding strategy data to achieve an adaptive and robust electricity bidding strategy.

[0041] In the embodiment of the present application, the distributed resources aggregated by VPP are as follows: micro gas turbine (MG), wind turbine (WT), photovoltaic station (PV), energy storage system (ESS), EVs and household loads (HL). Among them, HL includes fixed load, transferable load and interruptible load, and uses 1-norm and ∞-norm to constrain the uncertainty set of electricity price with incomplete probability distribution.

[0042] In this embodiment, the virtual power plant (VPP) aggregates multiple distributed energy resources to optimize power dispatch and bidding strategies. The micro gas turbine (MG) has the ability to start and stop quickly, and is used to provide stable power output, especially as a backup power source when renewable energy fluctuates greatly. The wind turbine (WT) uses wind energy to generate electricity. The uncertainty of wind speed makes its output volatile, which is suitable for use in combination with other stable power sources. Photovoltaic units (PV) use solar energy to generate electricity, which is greatly affected by weather and light intensity and has daytime peak power generation characteristics. The energy storage system (ESS) is used to store excess electricity, which can release electricity when electricity demand peaks or prices are high, balance electricity supply and demand, and provide flexible dispatch capabilities. Electric vehicles (EVs) are not only mobile loads, but also energy storage devices that feed electricity back to the grid and participate in demand response. Household loads (HL) include three types of loads: fixed loads, that is, the basic electricity demand that users cannot change on a daily basis. Transferable loads, loads that can be scheduled in different time periods, such as washing machines or charging equipment. Interruptible loads, loads that can be temporarily interrupted when electricity is tight or electricity prices are high, such as air conditioners and other equipment.

[0043] In order to cope with the uncertainty of electricity prices, the 1-norm and ∞-norm are used in the embodiments of the present application to constrain the incomplete probability distribution of electricity prices: 1-norm: used to measure the average deviation in the uncertain electricity price set, reflecting the overall range of electricity price fluctuations. Through this constraint, the overall risk brought by electricity price uncertainty can be controlled. ∞-norm: used to limit the worst case in the electricity price set, that is, the maximum possible deviation value of the electricity price, to ensure that the scheduling scheme is still robust in the most extreme cases. Through these norm constraints, VPP can maintain a stable scheduling and bidding strategy when dealing with electricity price fluctuations, reducing the impact of uncertainty on economic benefits.

[0044] In the embodiment of the present application, the load fuzzy response mechanism model refers to guiding users to transfer loads based on their fuzzy response behavior to time-of-use electricity prices, thereby optimizing the bidding strategy of virtual power plants (VPPs). This model takes into account the psychological reaction of users to electricity price changes and their uncertainty, and models and analyzes them through fuzzy logic.

[0045] Optionally, the load fuzzy response mechanism model is constructed based on a logistic algorithm. Specifically, the embodiment of the present application utilizes a load fuzzy response mechanism model based on logistic to guide users to transfer loads according to time-of-use electricity prices, and participates in a two-stage adaptive robust optimization framework as a bidding basis for VPP.

[0046] Among them, the stages of the load fuzzy response mechanism model are divided into: dead zone, response zone, and saturation zone. Different stages have different response characteristics and calculation methods. Further optionally, the user response behavior in the dead zone is determined by the average of the optimistic and pessimistic response predictions. This means that the model takes into account the possible responses of users in extreme cases (optimistic and pessimistic), and takes the average value to reflect the random response behavior of users when the electricity price difference is small. The electricity price difference is small, the user is not sensitive to changes in electricity prices, and the response behavior has greater randomness. In this area, since the electricity price difference is not enough to motivate users to significantly change their electricity consumption behavior, the user response is low. Further optionally, the user response behavior in the response zone is calculated by taking the optimistic response membership as the probability constraint of the demand response mechanism and using a biased semi-trapezoidal membership function. The optimistic response membership reflects the probability that the user responds to the optimistic estimate under the electricity price difference. The user response is calculated using a biased semi-trapezoidal membership function. As the electricity price difference increases, users become sensitive to changes in electricity prices and their response enthusiasm increases. The user's response behavior tends to be more optimistic, that is, they are more willing to reduce electricity consumption during peak electricity prices and increase electricity consumption during valley hours. Further optionally, the user's response behavior in the saturated area is determined based on the maximum value of the load transfer rate. When the price difference reaches a certain level, the user's responsiveness to changes in electricity prices reaches saturation, and no matter how large the price difference is, it will not significantly increase the user's responsiveness. Users have fully responded to changes in electricity prices in this area, and their load transfer rate has reached the maximum value. The user's response behavior is directly based on the maximum value of the load transfer rate, which means that the user's response behavior is no longer affected by further changes in the price difference.

[0047] In the embodiment of the present application, before participating in the multi-power market bidding, the VPP shall consider the effect of the time-of-use electricity price on the customers it serves, and make effective adjustments based on the actual response behavior of the users. According to consumer psychology, excessive or small price differences and the smoothness of the responsiveness curve at the point of price changes will cause changes in user demand response. Under the electricity price incentive mechanism, users are based on the principle of voluntariness, and their response behavior has obvious randomness. The actual demand response curve will be between the optimistic and pessimistic response prediction curves, and have fuzzy properties.

[0048] In the embodiment of the present application, taking peak-valley load transfer as an example, a fuzzy response mechanism model based on logistic function is established, such as Figure 2 As shown in the figure, λ PV is the load transfer rate, which can be used to describe the corresponding values ​​of different electricity price differences and different response curves, that is, the corresponding part of the curve in the figure; m is the optimistic response membership, which is used to describe the probability that the user meets the optimistic response estimate, that is, the area enclosed by the large semi-trapezoidal membership function in the bar chart; is the electricity price difference; a pv and b pvSet regional demarcation points for electricity price differences.

[0049] This model is divided into dead zone, response zone and saturation zone. In the dead zone, the price difference is too small and the user response behavior is highly random, so it is determined by the average of the optimistic and pessimistic response predictions. In the response zone, as the price difference increases, the user's response enthusiasm is mobilized, and the response behavior is more inclined to the optimistic response curve. Therefore, the optimistic response membership is considered and used as the probability constraint of the demand response mechanism. The large semi-trapezoidal membership function is used for calculation. In the saturation zone, the two curves coincide, and the user response behavior can be expressed by the maximum value of the load transfer rate. In this way, the load fuzzy response mechanism model improves the prediction accuracy of the user's demand response behavior through the division of three areas and different calculation methods, thereby optimizing the bidding strategy of the virtual power plant.

[0050] As an optional embodiment, a load fuzzy response mechanism model is adopted in 101 to generate load transfer parameters according to the time-of-use electricity price obtained in real time. Specifically, in step 101, using the load fuzzy response mechanism model to generate load transfer parameters according to the time-of-use electricity price obtained in real time is a key link in the adaptive robust bidding strategy of the virtual power plant. The model aims to dynamically predict the load transfer behavior of users through changes in time-of-use electricity prices, generate corresponding load transfer parameters (i.e., load response rate), and guide the decision-making of virtual power plants in bidding. The virtual power plant obtains the time-of-use electricity price in the current time period through the real-time market system. The time-of-use electricity price usually fluctuates with changes in market supply and demand and power generation costs, so it is an important factor affecting the user's electricity consumption behavior. The load fuzzy response mechanism model predicts how users will adjust their electricity load according to changes in electricity prices by considering the fuzzy response behavior of users to changes in electricity prices. The load transfer parameter represents the proportion of users adjusting their electricity load at the current electricity price, and the parameter is predicted based on the user's response to the time-of-use electricity price. Finally, the load fuzzy response mechanism model generates load transfer parameters based on the real-time time-of-use electricity prices and the response characteristics of users under different electricity prices. This parameter is an important basis for virtual power plants to adjust power distribution and make decisions in bidding.

[0051] In this way, the generated load transfer parameter represents the proportion of load transferred by users under the current electricity price conditions, which is used to optimize the bidding strategy of the virtual power plant. Based on this parameter, the virtual power plant can decide the load to be reduced when the electricity price is peak, or the load to be increased when the electricity price is low, so as to achieve the goal of reducing electricity costs and improving economic benefits. This dynamic load transfer mechanism enhances the flexibility and robustness of the virtual power plant in dealing with electricity price fluctuations.

[0052] As an optional embodiment, obtaining the bidding basis based on the load transfer parameters in 101 can be implemented as follows: performing a demand response result after load transfer based on the load transfer parameters; taking the demand response result as a reference, dynamically adjusting the load transfer parameters to obtain the final target load transfer parameters as the bidding basis.

[0053] Specifically, first, the initial load transfer parameters are generated based on the load fuzzy response mechanism model, and the preliminary results of the user's demand response are obtained by executing the corresponding load transfer operations. The demand response results reflect the actual response of users to load adjustment under the current electricity price conditions. Next, based on the preliminary demand response results, evaluate whether the effect of load transfer meets expectations. If the expected goal is not achieved, the initial load transfer parameters are dynamically adjusted, and the load transfer strategy is corrected considering the difference between the actual response of the user and the model prediction. Finally, through multiple iterative corrections, the target load transfer parameters that meet the actual situation are generated.

[0054] In this way, the target load transfer parameter is used as the bidding basis for the virtual power plant to make power allocation decisions in the market bidding. This can ensure that the bidding strategy is more accurate and improve economic benefits.

[0055] In the embodiment of the present application, further optionally, the day-ahead market bidding strategy model and the real-time market bidding strategy model are both adaptive robust optimization models.

[0056] Based on the above introduction, the least binomial support vector machine (LSSVM) is used as the driving model for data prediction, and the adaptive kernel density estimation (AKDE) is combined to generate a renewable energy uncertainty set, and the renewable energy uncertainty set is embedded into at least one of the above-mentioned adaptive robust optimization bidding models as original data.

[0057] In order to get rid of the overly conservative nature of the unipolar uncertainty set and overcome the huge workload caused by the pre-statistical prediction of data distribution errors, the embodiment of the present application uses the LSSVM-AKDE non-parametric estimation model to generate the uncertainty set. The least squares support vector machine is an improved algorithm of the support vector machine. It converts the inequality constraints in the SVM algorithm into equality constraints, and uses the least squares linear system as the loss function instead of the quadratic programming method used by the support vector machine. Its basic principle is expressed as follows:

[0058] A sample of training data can be represented as x i is the input value of the i-th sample, y i is the output value of the i-th sample, and l is the number of training samples. The LSSVM model in the feature space is shown in formula (1):

[0059]

[0060] (1) Where: is a nonlinear transformation mapping function that maps the input sample data to a high-dimensional feature space; ω is a weight vector; and b is a bias.

[0061] The objective function of LSSVM is shown in formula (2):

[0062]

[0063] (2) In the formula, ξ is the error variable; γ>0 is the penalty coefficient.

[0064] By introducing the Lagrangian function, we can solve formula (3):

[0065]

[0066] (3)In the formula, a is the Lagrange multiplier.

[0067] According to the KKT solution formula, calculate in sequence The following linear equations (4) are obtained:

[0068]

[0069] (4) In the formula, 1=[1,1,…,1] T ; q=[y1,y2,…,y l ] T ; a=[a1,a2,…,a l ] T ; I is the unit matrix.

[0070] According to the Mercer condition, the kernel function can be written as formula (5):

[0071]

[0072] After solving a and b by combining formula (4) and formula (5), we can get the nonlinear function of LSSVM:

[0073]

[0074] Different kernel functions in formula (6) generate different support vector bases. The most commonly used kernel functions are polynomial kernel function, linear kernel function, and radial basis kernel function (RBF). Since the RBF kernel function has a simple structure and strong generalization ability, it requires the least parameters to be optimized, only the kernel parameter σ and the regularization parameter γ, so this paper uses RBF.

[0075] KDE is based on the density function of the sample itself to determine the error and calculate the confidence interval. If the density function of the random variable X is f(x) and the empirical distribution function is F(x), then the simple estimate of f(x) is shown in formula (7):

[0076]

[0077] Where: {x (j) =1,2,...,Z} are sample points for constructing density function, n is the dimension of variable space, w is the window width parameter, K(·) is the kernel probability density function, which is usually a normal kernel density function, and its form is shown in formula (8):

[0078]

[0079] In the formula, S is the sample set {x (j)}, Used to measure each sample point x (j) The data dispersion in different directions and ranges. The window width parameter w controls the smoothness of the kernel density function. Too large w will cause the kernel density function to be too smooth, and too small w will cause errors in the tail of the kernel density function. AKDE is obtained by modifying the window parameter w on the basis of fixed bandwidth, and its form is as follows:

[0080]

[0081] In the formula, λ j is the local bandwidth factor, w′ is the window parameter corresponding to AKDE, which is obtained by formula (10-11):

[0082]

[0083] In the above formula, 0≤δ≤1 is the sensitivity factor, δ=0 corresponds to a fixed bandwidth, usually δ=0.5, Z d is the number of different samples. On this basis, the confidence interval under a certain confidence level can be calculated. When the confidence level is given as 1-α, if the sample value satisfies formula (12):

[0084]

[0085] In this case, is called the confidence interval of the sample value, where and represent the lower and upper limits of the interval respectively.

[0086] First, data preprocessing is performed to standardize the original data and map the data to the interval [-1,1] to eliminate the influence of different units and magnitudes on model training. Next, a multivariate regression model is established using the least squares support vector machine (LSSVM), in which the radial basis kernel function is used as the kernel function, and the penalty parameter and kernel function parameters are determined by cross-validation to ensure the generalization ability and prediction accuracy of the model. Then, the adaptive kernel density estimation (AKDE) method is used to estimate the residual distribution of the residual generated by the LSSVM model to accurately characterize the probability distribution of the prediction error. Finally, based on the preferences of EMS (energy management system) decision makers, the corresponding confidence interval parameters are set to construct a variable set of photovoltaic and wind power uncertainties, thereby providing interval predictions. This process, by combining LSSVM and AKDE, can effectively deal with the uncertainty of photovoltaic and wind power and improve the reliability and practicality of predictions.

[0087] For example, in a bilateral open electricity market trading model, VPP can choose to participate in bidding as a power generator or user. Based on the capacity limitation of VPP, the embodiment of the present application considers VPP participating in the trading activities of the electricity market as a price receiver, and the quotation of VPP will not affect the market clearing results. When VPP acts as a price receiver, it can submit a lower electricity selling price and a higher electricity purchasing price to the market operator (MO), thereby ensuring that all the bid electricity (sales or purchase) is successful, and VPP only needs to optimize its own bid electricity according to the predicted electricity market price.

[0088] As an optional embodiment, in 102, the first predicted bidding strategy data for the second trading day is predicted based on the historical trading data by using the day-ahead market bidding strategy model, which can be implemented as follows:

[0089] The day-ahead market bidding strategy model is used to calculate a first bidding cost corresponding to the historical transaction data; the first bidding cost is the sum of the DAM bidding cost, the micro gas turbine power generation cost, the IBDR project compensation cost, and the battery loss cost of the electric vehicle calculated based on the historical transaction data;

[0090] The first bidding cost is optimized through multiple first constraints; wherein the multiple first constraints include: at least one of bilateral contract constraints, power purchase and sales constraints of DAM and RTM, output constraints and ramp power limits of micro gas turbines, user transferable load and interruptible load constraints, electric vehicle charging and discharging restrictions, energy storage device charging and discharging restrictions, renewable energy consumption constraints, and power balance constraints;

[0091] The renewable energy uncertainty set is used to iteratively optimize and solve the optimized first bidding cost until a preset first convergence condition is met; the preset first convergence condition is at least that the difference between the first bidding costs before and after iterative optimization is less than a set threshold.

[0092] Before DAM is turned on, VPP guides users to make load transfers based on time-of-use electricity prices according to the fuzzy response mechanism of the logistic function, and uses the results after demand response as the total amount of bidding in multiple markets. VPP signs a contract with MO in the bilateral contract market to complete the preliminary trading plan. At 10 o'clock on the day before the trading day, the DAM market opens. At this time, the wind power and photovoltaic output, DAM and RTM electricity prices are unknown. EMS combines historical data with its own preferences to formulate corresponding bidding strategies and submit a day-ahead bidding plan to MO. The next day, VPP needs to complete the DAM transaction according to the submitted bidding electricity. At 13:00 on the day before the trading day, the DAM market closes and MO clears the DAM electricity price. Considering that the bidding power of DAM and RTM must meet the power balance constraint, the bidding power of VPP in RTM will affect the bidding power of DAM. Therefore, VPP formulates the bidding strategies of DAM and RTM at the same time, but only needs to submit the bidding strategy of DAM. The objective function of its bidding model is divided into four parts, as shown in formula (13). The first part represents the bidding cost of DAM, the second part represents the power cost of micro gas turbine, the third part represents the compensation cost of IBDR project, and the last part represents the battery loss cost of electric vehicle.

[0093]

[0094]

[0095]

[0096] Where, formula (14) represents the bidding cost of DAM, With D CON Sign bilateral contracts on behalf of VPP for electricity prices and volumes, and represents the predicted DAM and RTM electricity prices, k B With k S Represents the power purchase coefficient and the power sales coefficient, corresponding to the ratio of the power purchase price of the VPP participating in the bidding to the power sales price and the market power price, P t DA and Represents the amount of electricity purchased and sold by DAM, P t RT and Represents the amount of electricity purchased and sold by RTM. The electricity quantities of the above-mentioned markets participating in the bidding are all second-stage variables.

[0097] Formula (15) represents the power cost of the gas turbine, P GT.t Represents the gas turbine output, which is also the first stage variable, a GT 、b GT With c GT Represent the cost coefficient, k GT with u GT Represents the startup cost and startup status of the gas turbine, n f Represents the amount of pollutant gas, Q gt represents the emission intensity of the fth pollutant gas from the gas turbine, V f With Y f They respectively represent the environmental value and fine magnitude of the f-th pollutant gas. The specific implementation standards are derived from the environmental value table of pollutants in China's power industry.

[0098] Formula (16) represents the compensation cost of the IBDR project, represents the incentive cost of the shiftable load, Represents the amount of transferable load, which is also the first stage variable. A positive value represents a reduction in load during period t, and a negative value represents an increase in load. represents the amount of transferable load in HL, represents the fuzzy response of user load based on logistic function before DAM is turned on. The relationship between the two corresponds to formula (23), n r Represents the interrupt level number, represents the compensation price for the rth level of interrupted load, Represents the rth level interruption load in period t, and is also the first stage variable.

[0099] (17) represents the battery loss cost of EV, n v is the number of EVs, is the battery purchase cost of the vth EV, is the number of charge and discharge cycles in the battery life cycle of the vth EV, is the battery capacity of the vth EV, is the discharge depth of the vth EV battery, and in the embodiment of the present application, the value is 100%. and are the charging power and discharging power of the vth car in time period t, both of which are second-stage variables. and is the charging efficiency and discharging efficiency of the vth car, and Represent the status flags of EV charging, discharging and traveling respectively, and are restricted by formula (31). is the driving distance of the vth EV in period t, k EVis the power consumed by the vth EV per unit driving distance.

[0100] Formula (18) limits the range of variation of the amount of bilateral contracts signed, where With D BC Indicates the scope of variation and the amount of the specified bilateral contract.

[0101] Formula (19) is the restriction on electricity purchasing and selling behaviors in DAM and RTM at time t.

[0102] Formula (20) is the limit of electricity purchase and sales in DAM and RTM.

[0103] Formula (21) is the output limit of MG, and formula (22) is the climbing power limit of MG.

[0104] Formula (24) limits the power of the load that can be transferred at time t, where Corresponding to the maximum transferable load of HL at time t.

[0105] Formula (25) limits the total amount of load that can be transferred by HL in one day.

[0106] Formula (26) limits the power of each level of interruptible load at time t, where It is the largest share of interruptible load at each level.

[0107] Formula (27) limits the interruption amount of each level of interruptible load at adjacent moments, and It is the maximum interruptible proportion of each level of interruptible load in adjacent time periods.

[0108] Formulas (28)-(34) are the constraints of EV, while formulas (35)-(41) are the constraints of ESS. It is worth noting that and That is, the charging and discharging power of ESS, which is also the first-stage variable.

[0109] Formulas (42) and (43) represent the power limits of WT and PV, where and represents the uncertainty set boundary generated by the nonparametric estimation model, and, considering the findings of Ruiz and Conejo, which include that “natural” means that the generation capacity is as low as possible and the demand load is as high as possible, the lower bound of the uncertainty set boundary is taken as the constraint object.

[0110] Formula (44) represents the VPP internal power balance constraint.

[0111] Further, in 103, based on the real-time transaction data on the second trading day, a real-time bidding strategy calculation is performed through a real-time market bidding strategy model to obtain second predicted bidding strategy data for adjusting the real-time bidding power in the bidding basis, which can be implemented as follows:

[0112] The real-time market bidding strategy model is used to calculate a second bidding cost of the real-time bidding electricity corresponding to the real-time transaction data; the second bidding cost is the sum of the RTM bidding cost, the micro gas turbine power generation cost, the IBDR project compensation cost, and the battery loss cost of the electric vehicle calculated based on the real-time transaction data;

[0113] The second bidding cost is optimized through multiple second constraints; wherein the multiple second constraints include: at least one of the interruptible load, bilateral contract constraints, DAM and RTM power purchase and sales constraints, micro gas turbine output constraints and ramp power limits, user transferable load and interruptible load constraints, electric vehicle charging and discharging limits, energy storage device charging and discharging limits, renewable energy consumption constraints, and power balance constraints;

[0114] The renewable energy uncertainty set is used to iteratively optimize and solve the optimized second bidding cost until a preset second convergence condition is met; the preset convergence condition is at least that the difference between the second bidding costs before and after iterative optimization is less than a set threshold.

[0115] On the trading day, the RTM market is opened period by period. At this time, the DAM electricity price and bidding power are known, and the wind power and photovoltaic output forecast values ​​are more accurate. EMS re-formulates and submits the RTM bidding strategy to MO period by period. Similarly, during the trading period, VPP needs to complete the RTM transaction according to the submitted bidding power, and MO clears the RTM electricity price after the RTM market closes. The bidding model for the second stage is:

[0116]

[0117]

[0118] (19)-(20),(22),(28)-(41)(56)

[0119]

[0120] In the above formula, They represent the bilateral contract signing amount, power purchase amount, power sales amount, clearing power price, MG output and transferable load based on incentives in DAM during period t. represents the fuzzy response of the user load based on the logistic function before RTM is turned on. The newly added constraint (55) limits the amount of interruptible load. and Corresponding to the real-time predicted values ​​of WT and PV output at period t, (59) updates the power balance.

[0121] Specifically, in this embodiment, the virtual power plant (VPP) comprehensively considers a variety of distributed energy resources and minimizes the total cost through a two-stage optimization method. The first stage of optimization mainly deals with variables that have low scheduling flexibility or must be determined in advance. The output of the micro gas turbine (MG output), because the gas turbine takes a certain amount of time to start and shut down, its output needs to be determined in advance. The interactive power of traditional energy storage devices, including the charging and discharging power of the battery energy storage system in different time periods, also needs to be planned in advance to balance the supply and demand of electricity during the day. Transferable loads refer to loads that can be scheduled in different time periods, including household appliances such as washing machines and water heaters. The use time of these loads can be adjusted within a certain range to match the use of electricity during cheap periods. Interruptible loads, loads that can be temporarily interrupted when electricity prices are high or there is a shortage of electricity, such as air conditioners, electric heating and other equipment. The start and stop times of these loads also need to be determined in advance.

[0122] The second phase of optimization mainly deals with variables with high dispatch flexibility to adapt to real-time changes in market prices and actual demand. The charging and discharging power of electric vehicles (EVs). As mobile energy storage devices, EVs have greater charging and discharging flexibility, which are dynamically adjusted according to real-time electricity prices and electricity demand. Multi-market bidding output, including power bidding in the day-ahead market (DAM) and the real-time market (RTM). Due to the uncertainty of market prices and electricity demand, these bids need to be adjusted according to real-time conditions.

[0123] The goal of the entire optimization process is to minimize the total cost. The total cost includes, but is not limited to: power generation costs, such as fuel costs for micro gas turbines. Energy storage costs, the cost of electricity for charging and discharging, and the depreciation cost of batteries. Load adjustment costs, the possible reduction in comfort or other costs caused by the adjustment of transferable and interruptible loads. Power market transaction costs, including power procurement costs and sales revenue in the DAM and RTM markets.

[0124] In the first stage, the output of micro gas turbines, the interactive power of traditional energy storage devices, the power of transferable loads and interruptible loads are determined. By predicting electricity demand and electricity prices, these variables are optimized to achieve preliminary total cost minimization. After the DAM and RTM markets are closed, the second stage begins. According to the changes in actual electricity prices and electricity demand, the charging and discharging power of electric vehicles and the multi-market bidding output are dynamically adjusted. This step further optimizes the total cost through real-time data and achieves more accurate scheduling and electricity trading. Through this two-stage optimization approach, VPP is able to minimize total costs in an uncertain market environment while ensuring flexibility and reliability of electricity supply.

[0125] As an optional embodiment, in 104, based on the second prediction bidding strategy data, the real-time electricity price transaction information in each time period in the first prediction bidding strategy data is dynamically adjusted to realize an adaptive and robust electricity bidding strategy, which can be realized as follows:

[0126] The first-stage variables are set according to the output of the micro gas turbine, the interactive power of the traditional energy storage device, and the output of the transferable load and interruptible load projects in the first predicted bidding strategy data; the second-stage variables are set according to the charging and discharging power of the electric vehicle and the multi-market bidding output in the second predicted bidding strategy data; with the goal of minimizing the total cost of electric output, the real-time electric output variables in each time period on the second trading day are optimized to obtain the real-time electricity price trading information corresponding to each time period.

[0127] As an optional embodiment, after dynamically adjusting the real-time electricity price trading information for each time period in the first predictive bidding strategy data based on the second predictive bidding strategy data in 104, the balance of electric power resources can be cleared in real time based on the bidding basis, the second predictive bidding strategy data and the first predictive bidding strategy data before the market closes on the second trading day.

[0128] In order to cope with the randomness and volatility of renewable energy output, a two-stage data-driven adaptive robust optimization model is further constructed based on the deterministic optimization model of the first stage. The overall solution framework is as follows: Figure 3As shown. In this model, the output of fossil fuel generators, the output of traditional energy storage batteries, and the output of transferable loads and interruptible loads based on incentive mechanisms are set as first-stage variables. These variables are decisions made before the actual output of photovoltaic and wind power is determined, and will not be adjusted with changes in the actual scenario. In addition, the state of charge of electric vehicles and the output of multi-market bidding are set as second-stage variables. After obtaining the set of photovoltaic and wind power outputs with confidence intervals, these variables make decisions based on actual conditions, thereby achieving real-time adjustments. Through this two-stage adaptive robust model, the system can flexibly adjust the scheduling plan in the face of uncertainty in wind and solar output, thereby improving the stability and economy of the overall scheduling.

[0129] The adaptive robust model considering the uncertainty of wind power and photovoltaic output is as follows:

[0130]

[0131] In the formula, f DAM.1 is the first stage scheduling cost, f DAM.2 is the second-stage dispatch cost under the kth electricity price scenario, P k is the probability of the kth electricity price scenario occurring, and K is the number of typical scenarios screened.

[0132] The constraints of the adaptive robust model considering the uncertainty of wind power and photovoltaic output are divided into two parts. The first part is the same as the constraints of the first-stage deterministic optimization model, including multi-market bidding constraints (Equations (18)-(20)), gas turbine output constraints ((21)-(22)), IBDR project constraints ((23)-(27)), electric vehicle constraints ((28)-(34)), energy storage device constraints ((35)-(41)), renewable energy consumption constraints ((42)-(43)), and power balance constraints (44). The second part is the constraints of the comprehensive norm. In theory, the feasible domain Ω of the initial probability value of the discrete electricity price scenario is an arbitrary range, but in order to make the electricity price distribution more in line with the actual situation, the embodiment of the present application establishes an uncertain probability distribution set Ω centered on the initial probability distribution and with 1-norm and ∞-norm as constraints, as shown in formula (63):

[0133]

[0134] In the formula, θ1, θ ∞ are the probability deviation values ​​of 1-norm and ∞-norm, α1, α ∞ are the uncertainty probability confidence of 1-norm and ∞-norm, P k0 is the initial probability value under the kth scenario, and M is the total number of historical samples.

[0135] In summary, the general expression of the day-ahead-real-time two-stage virtual power plant adaptive robust bidding model is as follows:

[0136]

[0137] Cy k ≤c (64-3)

[0138] Gx+Hy k =g (64-4)

[0139] In the formula, is the first-stage variable, is the second stage variable, x * is the optimal solution of the variables in the first stage, s k is the set of uncertain probability distributions under the kth scenario constrained by the 1-norm and the ∞-norm, Q is the distribution of x * and k The feasible domain is constructed, and A, B, C, G, H, a, b, c, and g are all coefficient matrices corresponding to the constraints. (79-2) is the related constraint of the variables in the first stage, corresponding to equations (21)-(27), (35)-(41); (79-3) is the related constraint of the variables in the second stage, corresponding to equations (18)-(20), (28)-(34); (79-4) is the coupling constraint of the variables in the first stage and the variables in the second stage, corresponding to equations (42)-(44).

[0140] The formula (79) is a min-max-min three-layer optimization problem. Different from the traditional robust optimization which only considers the scheduling strategy under the worst case, the model proposed in the embodiment of the present application is dedicated to solving the decision variable {P k}, get the worst case of K typical scenarios, so as to obtain the maximum expected cost. The column and constraint generation algorithm (C&CG) can solve the above problem and obtain a finite convergence solution. The idea behind C&CG is to decompose the problem into a main problem (MP) and a subproblem (SP). MP solves the optimal decision of the first-stage variables that meet all constraints under the worst probability distribution identified by SP, and provides the lower bound of the model The main question is as follows:

[0141]

[0142] The sub-problem is to identify the worst probability distribution after solving the first-stage variables in the main problem, provide it to the main problem for further iterative calculation, and provide the upper bound of the model. The sub-questions are as follows:

[0143]

[0144] Since the first-stage variables of the inner optimization in subproblem (66) and the information of typical electricity price scenarios are known, there is no coupling relationship between the inner and outer optimization variables in the constraints, and there is no need to perform the primal-dual transformation on the subproblem. Therefore, the original problem can be decomposed into two models shown in Equations (67) and (68) and solved separately.

[0145]

[0146] In the formula, is the optimal value of the second-stage variable obtained by formula (68). When the difference between the optimized values ​​of (67) and (68) is less than a certain allowable convergence accuracy ε, the iteration stops and the original problem is solved.

[0147] The objective function of the original problem (64) is in linear form, while (63) contains absolute value constraints, which require equivalent linear transformation. When performing equivalent transformation on the absolute value constraint of the 1-norm, a 0-1 auxiliary variable is introduced and Represent the probability P k Relative to P k0 The 0-1 identifier of the positive and negative offsets. Considering that the offset state is unique, the following constraints are added:

[0148]

[0149] Therefore, formula (63) can be equivalently expressed as:

[0150]

[0151] The ∞-norm is handled in a similar way, and the constraint (71) is also introduced. However, unlike the 1-norm, the following constraints need to be added:

[0152]

[0153] At this point, the day-ahead and real-time two-stage adaptive robust bidding model for virtual power plants has been solved.

[0154] To verify the validity of the above model, this application example selects a VPP consisting of a gas turbine, an energy storage device, a wind turbine, a photovoltaic unit and 4,000 households. The pollutants emitted by the micro gas turbine are set to CO2, CO, SO2, NO x The emission quantity, environmental value and fine magnitude are shown in Table 1.

[0155] Table 1 Micro gas turbine pollutant gas parameters

[0156]

[0157] Among them, T = 24, the source load output in each time period is as follows Figure 4 The historical output data sets of wind power and photovoltaic power are from the first wind turbine and photovoltaic unit in a renewable energy fleet in a province in southern China, respectively. 4,000 sets of electricity prices from the day-ahead market and real-time market in Henan Province, China are selected and clustered into four typical scenarios. The specific electricity prices and initial probability distributions are shown in Figure 5 , where the peak and valley periods of various electricity prices follow the notice issued by the Henan Provincial Development and Reform Commission in 2022. To distinguish the purchase price from the sales price, the purchase coefficient is set to 1.1 and the sales price is set to 0.9. In the balanced market, the purchase price and sales price are 1.3 and 0.8 times the real-time market clearing price, respectively. Assuming that 4,000 households own 1,000 EVs of Nissan LEAF, BYD E6, Mitsubishi iMi and BMW MINIE types, the parameters of EVs and their travel patterns are shown in Table 2, and other equipment parameters are shown in Table 3.

[0158] Table 2 EV parameters

[0159]

[0160] Table 3 Main operating parameters of VPP

[0161]

[0162] The first set of historical data on wind power and photovoltaic output in the region from September 1 to September 10 was used as their respective training samples, and the data on September 11 was used as the test set. The corresponding results of different confidence intervals generated are shown in Table 4-5.

[0163] Table 4 Wind power output sets with different confidence intervals

[0164]

[0165]

[0166] Table 5 Photovoltaic output sets with different confidence intervals

[0167]

[0168]

[0169] In order to illustrate the quality of the interval estimation results, the embodiment of the present application refers to the prediction interval average interval width (PIAIW) and prediction interval coverage probability (PICP) parameters, and the results are shown in Table 6.

[0170] Table 6 Evaluation of prediction results parameters with different confidence intervals

[0171]

[0172] Analysis of wind power and photovoltaic data under different confidence intervals shows that as the confidence interval increases, the more information contained in the uncertainty set, the resulting focusing ability decreases, and the PINAW value increases. In addition, the PINAW values ​​of the confidence interval of 10% are 2.00% and 0.8%, respectively, which also shows that under conditions that are infinitely close to reality, the model proposed in the embodiment of the present application has good information focusing capabilities. As for the PICP value, the model proposed in the embodiment of the present application is higher than the corresponding confidence interval in most cases, indicating that the model proposed in the embodiment of the present application effectively completes the task of capturing interval information.

[0173] Through example simulation, the fuzzy response results of HL based on logistic function proposed in the embodiment of the present application are as follows Figure 6 As shown, the IBDR results are as follows Figure 7 As shown in Figure 2, the EV travel results are as follows: Figure 7 In order to verify the advantages of the model proposed in this specification in balancing economy and robustness when dealing with uncertainty, the following comparative cases are adopted. The optimization results of different schemes are shown in Table 7. The bidding results and power balance of VPP under different models are shown in Figure 8 shown.

[0174] Table 7 Optimization cost of different benchmarks

[0175]

[0176]

[0177] Scheme 1 adopts the traditional robust optimization strategy, selects the extreme electricity prices that appear in the clustering process as the electricity price uncertainty set, and generates the renewable energy uncertainty set based on the maximum deviation rate in the training set. Scheme 2 uses the model proposed in the embodiment of the present application, and sets the confidence interval of the uncertain set to 0.95. Scheme 3 is based on Scheme 2, but adjusts the confidence interval of the uncertain set to 0.90, and other conditions remain unchanged. Scheme 4 uses a two-stage random optimization strategy to model the problem. The electricity price group consists of clustering results with equal probability distribution, and optimizes the bidding process based on the predicted renewable energy output.

[0178] Under the strategy proposed in the embodiment of the present application, the total operating cost of the VPP is between Scheme 1 and Scheme 4, which is lower than Scheme 1 using the traditional robust optimization strategy, and higher than Scheme 4 using two-stage stochastic optimization. Specifically, compared with Scheme 1, the scheme of the embodiment of the present application effectively utilizes the initial distribution information of the electricity price and improves the prediction accuracy of renewable energy, overcoming the disadvantage of Scheme 1 being too conservative, so the total cost of the VPP is reduced. Compared with Scheme 2, the two-stage stochastic optimization strategy is based on the assumption that the electricity price obeys an equiprobable probability distribution. It does not consider the impact of the worst-case probability distribution of uncertainty on the VPP decision. Therefore, the total cost is expected to be lower than the total cost in Scheme 2. In addition, Scheme 3 reduces the conservatism of the confidence interval, making the initial uncertainty set considered closer to the actual situation, thereby reducing the optimization cost. In short, the model proposed in the embodiment of the present application takes into account both economy and robustness under a smaller computational burden and higher efficiency.

[0179] The above simulation results verify the effectiveness and practicality of the method of the embodiment of the present application. The method of the embodiment of the present application effectively combines the limited rational response results of residential users to time-of-use electricity prices with distributed energy in the two-stage bidding process of VPP day-ahead and real-time, and in the face of the challenge of uncertain output of renewable energy, the proposed uncertainty set can effectively concentrate the uncertainty information on a narrow boundary, and the AKDE curve based on the 95% confidence interval can balance accuracy and selectivity while retaining the original PDF information. Compared with the traditional robust model and the two-stage stochastic optimization model, this strategy can cope with more practical scenarios according to the preferences of the EMS, and has better economy and robustness.

[0180] In the embodiment of the present application, a load fuzzy response mechanism model is adopted to generate load transfer parameters according to the time-of-use electricity price obtained in real time, and obtain the bidding basis based on the load transfer parameters; obtain the historical transaction data of the first trading day, and use the day-ahead market bidding strategy model to predict the first predicted bidding strategy data within the second trading day based on the historical transaction data; based on the real-time transaction data of the second trading day, the real-time bidding strategy is calculated through the real-time market bidding strategy model to obtain the second predicted bidding strategy data; based on the second predicted bidding strategy data, the real-time electricity price transaction information in each time period is dynamically adjusted to realize an adaptive and robust electricity bidding strategy. Through the load fuzzy response mechanism and the adaptive and robust optimization model, the bidding risk is reduced, and the economic benefits of the virtual power plant and the stability of the power market are improved.

[0181] In another embodiment of the present application, an adaptive robust bidding system for a virtual power plant is also provided. Fig. 9 The system comprises the following units:

[0182] A generating unit is configured to adopt a load fuzzy response mechanism model, generate a load transfer parameter according to a time-of-use electricity price obtained in real time, and obtain a bidding basis based on the load transfer parameter; the bidding basis is used to indicate the total amount of bidding to participate in multiple markets;

[0183] The first prediction unit is configured to obtain historical transaction data of the first trading day, and predict first predicted bidding strategy data within the second trading day based on the historical transaction data through a day-ahead market bidding strategy model; the historical transaction data includes at least one of the following: historical wind power output data, historical photovoltaic output data, historical market electricity price data, and historical user load change data; the first predicted bidding strategy data includes at least one of the following: gas turbine output, energy storage device power, and output of transferable and interruptible load projects;

[0184] The second prediction unit is configured to perform a real-time bidding strategy calculation based on the real-time transaction data of the second trading day through a real-time market bidding strategy model to obtain second predicted bidding strategy data for adjusting the real-time bidding power; the second predicted transaction data includes at least one of the following: electric vehicle charging and discharging power, and multi-market bidding output;

[0185] The real-time bidding unit is configured to dynamically adjust the real-time electricity price transaction information in each time period in the first predictive bidding strategy data based on the second predictive bidding strategy data to achieve an adaptive and robust electricity bidding strategy.

[0186] Further optionally, the day-ahead market bidding strategy model and the real-time market bidding strategy model are both adaptive robust optimization models; the system further includes a generation module configured to:

[0187] The least binomial support vector machine (LSSVM) is used as a driving model for data prediction, and is combined with an adaptive kernel density estimation (AKDE) to generate a renewable energy uncertainty set, which is then embedded into at least one of the above-mentioned adaptive robust optimization bidding models as raw data.

[0188] Further optionally, the first prediction unit predicts the first predicted bidding strategy data within the second trading day based on the historical trading data through the day-ahead market bidding strategy model, and is configured to:

[0189] The day-ahead market bidding strategy model is used to calculate a first bidding cost corresponding to the historical transaction data; the first bidding cost is the sum of the DAM bidding cost, the micro gas turbine power generation cost, the IBDR project compensation cost, and the battery loss cost of the electric vehicle calculated based on the historical transaction data;

[0190] The first bidding cost is optimized through multiple first constraints; wherein the multiple first constraints include: at least one of bilateral contract constraints, power purchase and sales constraints of DAM and RTM, output constraints and ramp power limits of micro gas turbines, user transferable load and interruptible load constraints, electric vehicle charging and discharging restrictions, energy storage device charging and discharging restrictions, renewable energy consumption constraints, and power balance constraints;

[0191] The renewable energy uncertainty set is used to iteratively optimize and solve the optimized first bidding cost until a preset first convergence condition is met; the preset first convergence condition is at least that the difference between the first bidding costs before and after iterative optimization is less than a set threshold.

[0192] The second prediction unit, based on the real-time transaction data of the second trading day, performs real-time bidding strategy calculation through the real-time market bidding strategy model to obtain second predicted bidding strategy data for adjusting the real-time bidding power in the bidding basis, and is configured as follows:

[0193] The real-time market bidding strategy model is used to calculate a second bidding cost of the real-time bidding electricity corresponding to the real-time transaction data; the second bidding cost is the sum of the RTM bidding cost, the micro gas turbine power generation cost, the IBDR project compensation cost, and the battery loss cost of the electric vehicle calculated based on the real-time transaction data;

[0194] The second bidding cost is optimized through multiple second constraints; wherein the multiple second constraints include: at least one of the interruptible load, bilateral contract constraints, DAM and RTM power purchase and sales constraints, micro gas turbine output constraints and ramp power limits, user transferable load and interruptible load constraints, electric vehicle charging and discharging limits, energy storage device charging and discharging limits, renewable energy consumption constraints, and power balance constraints;

[0195] The renewable energy uncertainty set is used to iteratively optimize and solve the optimized second bidding cost until a preset second convergence condition is met; the preset convergence condition is at least that the difference between the second bidding costs before and after iterative optimization is less than a set threshold.

[0196] Further optionally, the real-time bidding unit dynamically adjusts the real-time electricity price transaction information in each time period in the first predicted bidding strategy data based on the second predicted bidding strategy data to implement an adaptive and robust electricity bidding strategy, and is configured as follows:

[0197] According to the output of the micro gas turbine, the interactive power of the traditional energy storage device, the output of the transferable load and the interruptible load project in the first prediction bidding strategy data, the first stage variables are set;

[0198] Setting second-stage variables according to the charging and discharging power of the electric vehicle and the multi-market bidding output in the second prediction bidding strategy data;

[0199] With the goal of minimizing the total cost of electricity output, the real-time electricity output variables in each period of the second trading day are optimized to obtain the real-time electricity price trading information corresponding to each period.

[0200] Further optionally, the load fuzzy response mechanism model is constructed based on a logistic algorithm;

[0201] Among them, the stages of the load fuzzy response mechanism model are divided into: dead zone, response zone, and saturation zone; the user response behavior in the dead zone is determined by the average value of the optimistic and pessimistic response predictions; the user response behavior in the response zone is calculated by taking the optimistic response membership as the probability constraint of the demand response mechanism and adopting a large semi-trapezoidal membership function; the user response behavior in the saturation zone is determined based on the maximum value of the load transfer rate.

[0202] Further optionally, the generating unit obtains the bidding basis based on the load transfer parameter and is configured to:

[0203] a demand response result after performing load transfer based on the load transfer parameter;

[0204] The load transfer parameters are dynamically adjusted with the demand response results as a reference to obtain the final target load transfer parameters as the bidding basis.

[0205] Further optionally, the real-time bidding unit, after dynamically adjusting the real-time electricity price transaction information in each time period in the first predicted bidding strategy data based on the second predicted bidding strategy data, is further configured to:

[0206] Before the market closes on the second trading day, the balance of electric power resources is cleared in real time based on the bidding basis, the second predicted bidding strategy data, and the first predicted bidding strategy data.

[0207] The system can implement various steps in the above method embodiment, which will not be expanded here.

[0208] In the embodiment of the present application, a load fuzzy response mechanism model is adopted to generate load transfer parameters according to the time-of-use electricity price obtained in real time, and a bidding basis is obtained based on the load transfer parameters; the historical transaction data of the first trading day is obtained, and the first predicted bidding strategy data within the second trading day is predicted based on the historical transaction data through the day-ahead market bidding strategy model; based on the real-time transaction data of the second trading day, the real-time bidding strategy is calculated through the real-time market bidding strategy model to obtain the second predicted bidding strategy data for adjusting the real-time bidding electricity in the bidding basis; based on the second predicted bidding strategy data, the real-time electricity price transaction information in each time period in the first predicted bidding strategy data is dynamically adjusted to realize an adaptive and robust electricity bidding strategy. In the embodiment of the present application, the bidding strategy is dynamically adjusted based on the time-of-use electricity price through the load fuzzy response mechanism and the adaptive robust optimization model to effectively respond to electricity price fluctuations and uncertainty in renewable energy output; historical transaction data and real-time data are used to optimize the bidding plan, reduce bidding risks, and improve the economic benefits of virtual power plants and the stability of the electricity market.

[0209] See also Fig.10 , Fig.10 This is a schematic diagram of an electronic device provided in an embodiment of the present application. Fig.10 As shown, an embodiment of the present application provides an electronic device 500, including a memory 510, a processor 5102, and a computer program 511 stored in the memory 510 and executable on the processor 5102. When the processor 5102 executes the computer program 511, the following steps are implemented: using a load fuzzy response mechanism model, generating load transfer parameters according to the time-of-use electricity price obtained in real time, and obtaining a bidding basis based on the load transfer parameters; obtaining historical transaction data of the first trading day, and using a day-ahead market bidding strategy model to predict first predicted bidding strategy data within the second trading day based on the historical transaction data; based on the real-time transaction data of the second trading day, calculating the real-time bidding strategy through the real-time market bidding strategy model, obtaining second predicted bidding strategy data for adjusting the real-time bidding electricity in the bidding basis; based on the second predicted bidding strategy data, dynamically adjusting the real-time electricity price transaction information in each time period in the first predicted bidding strategy data to achieve an adaptive and robust electricity bidding strategy.

[0210] See also Fig.11 , Fig.11 A schematic diagram of an embodiment of a computer-readable storage medium provided in an embodiment of the present application. Fig.11As shown, this embodiment provides a computer-readable storage medium 600 on which a computer program 611 is stored. When the computer program 611 is executed by a processor, the following steps are implemented: using a load fuzzy response mechanism model, generating load transfer parameters according to the time-of-use electricity price obtained in real time, and obtaining a bidding basis based on the load transfer parameters; obtaining historical transaction data of the first trading day, and using a day-ahead market bidding strategy model to predict first predicted bidding strategy data for the second trading day based on the historical transaction data; based on the real-time transaction data of the second trading day, performing real-time bidding strategy calculations through a real-time market bidding strategy model to obtain second predicted bidding strategy data for adjusting the real-time bidding electricity in the bidding basis; based on the second predicted bidding strategy data, dynamically adjusting the real-time electricity price transaction information in each time period in the first predicted bidding strategy data to achieve an adaptive and robust electricity bidding strategy.

[0211] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0212] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0213] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0214] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0215] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0216] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present application.

[0217] Obviously, those skilled in the art can make various changes and modifications to the embodiments of the present application without departing from the spirit and scope of the embodiments of the present application. Thus, if these modifications and variations of the embodiments of the present application fall within the scope of the claims of the embodiments of the present application and their equivalents, the embodiments of the present application are also intended to include these modifications and variations.

Claims

1. An adaptive robust bidding method for a virtual power plant, characterized in that: The method comprises: A load fuzzy response mechanism model is adopted to generate load transfer parameters according to the time-of-use electricity price obtained in real time, and a bidding basis is obtained based on the load transfer parameters; the bidding basis is used to indicate the total amount of bidding in multiple markets; Obtain historical transaction data of the first trading day, and predict first predicted bidding strategy data for the second trading day based on the historical transaction data through a day-ahead market bidding strategy model; the historical transaction data includes at least one of the following: historical wind power output data, historical photovoltaic output data, historical market electricity price data, and historical user load change data; the first predicted bidding strategy data includes at least one of the following: gas turbine output, energy storage device power, and output of transferable and interruptible load projects; Based on the real-time transaction data of the second trading day, a real-time bidding strategy is calculated by a real-time market bidding strategy model to obtain second predicted bidding strategy data for adjusting the real-time bidding power in the bidding basis; the second predicted transaction data includes at least one of the following: electric vehicle charging and discharging power, and multi-market bidding output; Based on the second predicted bidding strategy data, the real-time electricity price transaction information in each time period in the first predicted bidding strategy data is dynamically adjusted to achieve an adaptive and robust electricity bidding strategy.

2. The adaptive robust bidding method for a virtual power plant according to claim 1, characterized in that: The day-ahead market bidding strategy model and the real-time market bidding strategy model are both adaptive robust optimization models; the method further includes: The least binomial support vector machine (LSSVM) is used as a driving model for data prediction, and is combined with an adaptive kernel density estimation (AKDE) to generate a renewable energy uncertainty set, which is then embedded into at least one of the above-mentioned adaptive robust optimization bidding models as raw data.

3. The adaptive robust bidding method for a virtual power plant according to claim 2, characterized in that: The method of predicting first predicted bidding strategy data for the second trading day based on the historical trading data by using the day-ahead market bidding strategy model includes: The day-ahead market bidding strategy model is used to calculate a first bidding cost corresponding to the historical transaction data; the first bidding cost is the sum of the DAM bidding cost, the micro gas turbine power generation cost, the IBDR project compensation cost, and the battery loss cost of the electric vehicle calculated based on the historical transaction data; The first bidding cost is optimized through multiple first constraints; wherein the multiple first constraints include: at least one of bilateral contract constraints, power purchase and sales constraints of DAM and RTM, output constraints and ramp power limits of micro gas turbines, user transferable load and interruptible load constraints, electric vehicle charging and discharging restrictions, energy storage device charging and discharging restrictions, renewable energy consumption constraints, and power balance constraints; The renewable energy uncertainty set is used to iteratively optimize and solve the optimized first bidding cost until a preset first convergence condition is satisfied; the preset first convergence condition is at least that the difference between the first bidding cost before and after iterative optimization is less than a set threshold; The real-time transaction data on the second trading day is used to calculate the real-time bidding strategy through the real-time market bidding strategy model to obtain the second predicted bidding strategy data for adjusting the real-time bidding power in the bidding basis, including: The real-time market bidding strategy model is used to calculate a second bidding cost of the real-time bidding electricity corresponding to the real-time transaction data; the second bidding cost is the sum of the RTM bidding cost, the micro gas turbine power generation cost, the IBDR project compensation cost, and the battery loss cost of the electric vehicle calculated based on the real-time transaction data; The second bidding cost is optimized through multiple second constraints; wherein the multiple second constraints include: at least one of the interruptible load, bilateral contract constraints, DAM and RTM power purchase and sales constraints, micro gas turbine output constraints and ramp power limits, user transferable load and interruptible load constraints, electric vehicle charging and discharging limits, energy storage device charging and discharging limits, renewable energy consumption constraints, and power balance constraints; The renewable energy uncertainty set is used to iteratively optimize and solve the optimized second bidding cost until a preset second convergence condition is met; the preset convergence condition is at least that the difference between the second bidding costs before and after iterative optimization is less than a set threshold.

4. The adaptive robust bidding method for a virtual power plant according to claim 1, characterized in that: Based on the second forecast bidding strategy data, the real-time electricity price transaction information in each time period in the first forecast bidding strategy data is dynamically adjusted to realize an adaptive and robust electricity bidding strategy, including: According to the output of the micro gas turbine, the interactive power of the traditional energy storage device, the output of the transferable load and the interruptible load project in the first prediction bidding strategy data, the first stage variables are set; Setting second-stage variables according to the charging and discharging power of the electric vehicle and the multi-market bidding output in the second prediction bidding strategy data; With the goal of minimizing the total cost of electricity output, the real-time electricity output variables in each period of the second trading day are optimized to obtain the real-time electricity price trading information corresponding to each period.

5. The adaptive robust bidding method for a virtual power plant according to claim 1, characterized in that: The load fuzzy response mechanism model is constructed based on the logistic algorithm; Among them, the stages of the load fuzzy response mechanism model are divided into: dead zone, response zone, and saturation zone; the user response behavior in the dead zone is determined by the average value of the optimistic and pessimistic response predictions; the user response behavior in the response zone is calculated by taking the optimistic response membership as the probability constraint of the demand response mechanism and adopting a large semi-trapezoidal membership function; the user response behavior in the saturation zone is determined based on the maximum value of the load transfer rate.

6. The adaptive robust bidding method for a virtual power plant according to claim 1, characterized in that: A bidding basis is obtained based on the load transfer parameter, including: a demand response result after performing load transfer based on the load transfer parameter; The load transfer parameters are dynamically adjusted with the demand response results as a reference to obtain the final target load transfer parameters as the bidding basis.

7. The adaptive robust bidding method for a virtual power plant according to claim 1, characterized in that: After dynamically adjusting the real-time electricity price transaction information in each time period in the first forecast bidding strategy data based on the second forecast bidding strategy data, the method further includes: Before the market closes on the second trading day, the balance of electric power resources is cleared in real time based on the bidding basis, the second predicted bidding strategy data, and the first predicted bidding strategy data.

8. An adaptive robust bidding system for a virtual power plant, characterized in that: The system comprises the following units, wherein: A generating unit is configured to adopt a load fuzzy response mechanism model, generate a load transfer parameter according to a time-of-use electricity price obtained in real time, and obtain a bidding basis based on the load transfer parameter; the bidding basis is used to indicate the total amount of bidding to participate in multiple markets; The first prediction unit is configured to obtain historical transaction data of the first trading day, and predict first predicted bidding strategy data within the second trading day based on the historical transaction data through a day-ahead market bidding strategy model; the historical transaction data includes at least one of the following: historical wind power output data, historical photovoltaic output data, historical market electricity price data, and historical user load change data; the first predicted bidding strategy data includes at least one of the following: gas turbine output, energy storage device power, and output of transferable and interruptible load projects; The second prediction unit is configured to perform a real-time bidding strategy calculation based on the real-time transaction data of the second trading day through a real-time market bidding strategy model to obtain second predicted bidding strategy data for adjusting the real-time bidding power; the second predicted transaction data includes at least one of the following: electric vehicle charging and discharging power, and multi-market bidding output; The real-time bidding unit is configured to dynamically adjust the real-time electricity price transaction information in each time period in the first predictive bidding strategy data based on the second predictive bidding strategy data to achieve an adaptive and robust electricity bidding strategy.

9. An electronic device, characterized in that: include: Memory for storing computer software programs; A processor is used to read and execute the computer software program, thereby implementing the adaptive robust bidding method for the virtual power plant as described in any one of claims 1-7.

10. A non-transitory computer-readable storage medium, characterized in that: The storage medium stores a computer software program, and when the computer software program is executed by the processor, it implements the adaptive robust bidding method for the virtual power plant as described in any one of claims 1-7.

Images